Workshop on Pickling Solutions Technology Workshop on Pickling Solutions Technology Optimisation of pickling process control and management by model-based simulation tools University of Oviedo Iván Machón González 13th of November 2019, Düsseldorf 1
Workshop on Pickling Solutions Technology Optimisation of pickling process control and management by model-based simulation tools Data correlations Data clustering • • Analysis and/or verification of Some algorithms can be used for cluster correlations analysis. Search of common patterns by means of merging similar samples. • Search of nonlinear or partial/local • correlations by means of visualization Classification tasks. algorithms 2
Workshop on Pickling Solutions Technology Optimisation of pickling process control and management by model-based simulation tools Data representation Machine Learning algorithms • Data representation (e.g. plots, • Supervised versus unsupervised barcharts, etc.) of results for further learning discussion with experienced personnel. 3
Workshop on Pickling Solutions Technology Optimisation of pickling process control and management by model-based simulation tools Development of the process model • Condition Monitoring and Predictive Maintenance General procedure 4
Workshop on Pickling Solutions Technology SensorControlPilot (I) • Estimation the pickling strip speed by means of a model that indicates the mean values and standard deviation of the maximum speed for different conditions in the pickling line. • Neural Gas network as model to establish a probabilistic distribution of the pickling line speed. The main idea is to calculate the optimum strip speed of the pickling line given the remaining process variables. Data from the hot rolling mill and the pickling line were used. • Two different trained models were considered depending on the material destination: chromium or tin. • The aim is to obtain a set of prototypes of coils by the application of this kind of algorithms. These prototypes are synthesizing all the information of the coils and they can be used to estimate the optimum strip speed of the pickling line. 5
Workshop on Pickling Solutions Technology SensorControlPilot (II) • The following process variables were used to train the neural model: – for tinned material destination: hot rolling coiling temperature, initial strip temperature, iron concentration in bath 1, acid concentration in bath 1, steel type, destination, strip thickness, strip width and pickling line speed. – for chromed material destination: hot rolling coiling temperature, destination, steel type, iron concentration in bath 1, acid concentration in bath 1, pickling line speed and strip thickness. • The euclidean distance within input data space for taking out the estimation of the strip speed setpoint. 6
Workshop on Pickling Solutions Technology MACOPilot Development of an innovative pickling program management model based on online data of the wire rod pickling plant process. • Specification of the pickling dwell time before the beginning of the pickling treatment by the management software tool. • Selection of the dip tank by means of the management model based on the current process data for optimal pickling result. • Simulated testing of the new pickling program management for wire rod pickling plant operation of DEW 7
Workshop on Pickling Solutions Technology Initial study of the variables affecting the effectiveness of the pickling process - Acid mixture composition and free Fe content. stable due to acid bath regenerations. - Steel type and allow composition. Fixed variables in the process - Previous heat treatment (austenization, annealing, tempering …). datasets (dictated by steel code) - Dwell time and number of consecutive pickling operations. - Hydraulic conditions in the acid baths. - Mixed acid bath temperature. The temperature dynamics in the pickling baths are increased during the treatment due to the combination of the pickling exothermic reaction and the cooling system refrigeration. The control of the temperature is essential for the development of the pickling: - Too low temperatures decrease the efficiency of the pickling reaction (poor treatment results): recommended to pickle over 25 ºC. - Too high temperatures affect the results of the pickling (more risk of overpickling and toxic steam emissions): higher temperature limit set at 40-45°C. Identification of the heat flux distribution due to: Necessary to develop a model of the temperature - Pickling exothermic reaction. dynamics which can predict its evolution. - Cooling system dynamics. 8
Workshop on Pickling Solutions Technology Development of the pickling programme management model TF models for the behaviour 10 -3 9 2 of the steels in the acid 1.8 8 exothermic reaction (pickling 1.6 7 process): heat flux ∁ ≡ 𝐼𝑓𝑏𝑢 𝑑𝑏𝑞𝑏𝑑𝑗𝑢𝑧 1.4 distribution identification and 6 𝑟 𝑜𝑓𝑢 𝑢 = ∁ 𝑒𝑈(𝑢) qsteel/C ( º C) 1.2 prediction for the pickling N º of coils º 5 𝑒𝑢 1 reaction. 4 º 0.8 3 Obtained by ARMAX 0.6 2 identification concerning bath 0.4 temperature datasets for wire 1 0.2 rod materials. 𝑟 𝑜𝑓𝑢 𝑢 = 𝑟 𝑡𝑢𝑓𝑓𝑚 𝑢 − 𝑟 𝑚𝑝𝑡𝑡 (𝑢) 𝑟 𝑚𝑝𝑡𝑡 𝑢 = (𝑈 𝑢 − 𝑈 0 ) 0 0 16:30 17:30 𝑆 𝑢 𝑆 𝑢 ≡ 𝑈ℎ𝑓𝑠𝑛𝑏𝑚 𝑠𝑓𝑡𝑗𝑡𝑢𝑏𝑜𝑑𝑓 Identification of the cooling Obtaining the heat flux distribution dynamics of the pickling baths (heat loss flux). corresponding to the temperature evolution as a result of the combination of the cooling system and pickling reactions. 9
Workshop on Pickling Solutions Technology Analysis of process variables influencing heat flux and temperature evolution Martensitic: special care (reaction triggered, steam emission, short dwell times). Important differences Austenitic and duplex: hardest to pickle, not important for temperature troubleshooting or overpickling. between each type of steel Ferritic: easiest to pickle, medium size dwell times. Noticeable differences The more alloy, the more TF model for each steel code between steels of each difficulty in pickling. category The amount of previous pickling stages TF model for each steel code in carried out affects the subsequent pickling each pickling stage operation. Reducing the shooting of the temperature (since a large part of the scale has already been eliminated previously). 10
Workshop on Pickling Solutions Technology Analysis of process variables influencing heat flux and temperature evolution II Fixed dwell times for each pickling programme Martensitic steels Exothermic reaction not finished before the coil is taken out Variable equivalent to 𝑅𝑡𝑢𝑓𝑓𝑚 𝑛𝑏𝑦 reaction speed Speed of chemical reaction 𝐷 Average speed 𝛿 = affects the heat flux produced by of the heating 𝐸𝑥𝑓𝑚𝑚 𝑢𝑗𝑛𝑓 the exothermic reaction of each pickling operation. 𝛿 Temperature evolution affected by the speed of the reaction. 𝛿 Temperature 𝑅𝑡𝑢𝑓𝑓𝑚 𝑛𝑏𝑦 triggering (for a fixed dwell time) 11
Workshop on Pickling Solutions Technology Analysis of process variables influencing heat flux and temperature evolution III • Variables affecting the speed of the chemical reaction due to the pickling process - Lower dwell times ( ≈ 3 min). • Temperature of the Higher dependence for - Reaction not completed when the coil is taken out of the bath. martensitic steels - Greater temperature triggering for higher bath temperature. bath (ºC) - Medium dwell times ( ≈ 8 min ). • Medium dependence for - Reaction almost completed before the coil is taken out of the bath ferritic steels (Qsteel max reached). Affects the speed of - Higher dwell times ( ≈ 15 min). • Lower dependence for temperature heating and - Reaction and temperature triggering completed before the coil is taken cooling austenitic/duplex steels out of the bath. 10 -3 B6 BATH heat flux after treatment 40 3 Final Temperature Evolution (after interpolation) Initial Threshold Temperature 39 2.5 Process number of coils (30+5*Nºcoils) qsteel/C 38 qloss/C 2 qnet/C 37 Arrhenius behavior 1.5 36 q/C ( º C) T( º C) 35 1 𝛿 34 𝛿 0.5 33 0 32 -0.5 31 30 -1 10:30 10:45 11:00 11:15 11:30 11:45 Mar 15, 2018 Heat flux distribution, two coils of ferritic steel 1.47420-02 (B6 tank, BP 40). Heat flux distribution, three coils of duplex steel 1.44620-54 (B6 tank) 12
Workshop on Pickling Solutions Technology Analysis of process variables influencing heat flux and temperature evolution IV • Variables affecting the contact surface area for the reaction • Considering approximately the same density for every steel type, they can be considered as the variables to evaluate the influence of the pickled surface in the reaction. 𝑊 ≈ 𝐸𝐵 𝐵 ≈ 𝜌𝐸𝑀 𝐵 ≈ 𝑙 1 Contact surface 𝑊 = 𝜌𝐸 2 4 𝐸 4 𝑀 Volume For coils of the same weight, 𝑊 1 ≈ 𝑊 2 Weight • The higher the weight, the (Kg) and higher the volume and the thickness greater the contact surface. (mm) of W ≈ 𝑊 ≈ 𝑙′𝐵 the coil 13
Recommend
More recommend
